Key takeaway
Heading density below 0.5 per 100 words leaves nothing to chunk on. Above 3.0, headings stop meaning anything. Both ends fail, which is the whole argument for bands over a score: "write better" becomes a number that is out of range.
Every page is fetched server-side and parsed, so the audit sees what a crawler sees rather than what a browser renders.
Get a list of what is missing, not a grade. Heading hierarchy, FAQ block, lists, structured data, a key takeaway, images, external links, statistics, an intro summary under 80 words, a table, a named author, freshness and authoritative sources, each present or absent with nowhere to hide.
Argue with the thresholds instead of accepting a verdict. Eleven bands each with a floor and a ceiling, paragraphs 30 to 90 words, list ratio 0.10 to 0.40, heading density 0.5 to 3.0 per 100 words and eight more, all published as numbers.
Get audited as the business you actually are. A lead-generation site is additionally checked for comparison content, testimonials, case studies and social proof; an ecommerce site for product schema, visible pricing and reviews.
Stop reconciling the audit against a separate crawl. Status code, indexability, robots directive, canonical, title, meta, H1 count, schema types and link counts come from the same pass.
A composite grade tells you a page is weak. A band tells you which number to change.
Heading depth, section count, heading density and whether a table of contents exists. This is the level that decides whether an engine can find the passage that answers a question, rather than the page as a whole.
Average and 90th-percentile paragraph length, list ratio and chunk variance. Uneven chunking is the failure mode a word count cannot see: a page can average 60 words a paragraph and still hide a 400-word wall.
Emphasis density, blockquotes, code density and image-to-text ratio. Under-emphasised text gives an extractor nothing to anchor on; over-emphasised text gives it too much and the signal flattens.
Each band has a floor and a ceiling because both ends fail. Too few headings and there is no structure; too many and every heading means less. The bands come from published research on generative-engine extraction and are re-tunable against observed citation rates.
Each crawl stores a snapshot. When a page changes, the audit reports which fields moved: title, meta description, H1, structured data, body copy. So a score change is always attributable to an edit.
The average page audit score across your audited pages is the Be Retrieved rung on the readiness ladder. One page's failure is a task; the average moving is a programme working.
The point of a numeric band is that it ends the argument about whether a page is good enough.
Stop telling a writer to write better when structure is the fault. Most pages that fail extraction are well written and badly structured, and a 400-word paragraph with no headings is not a quality problem.
Check a draft before you ship it. The bands are computable on unpublished content, so you never have to publish, wait a fortnight and infer from citation counts whether structure was the problem.
Prove the rewrite worked. A page change becomes a change event with a field-level diff, and the prompts it should have moved become the treatment group against a control drawn from those it should not.
The difference is whether the output names a number you can change.
Published numeric bands with floors and ceilings
TrustData
GA4
Other tools
Server-side fetch, sees what a crawler sees
TrustData
GA4
Other tools
SometimesSignals differ by site type
TrustData
GA4
Other tools
Field-level diff between crawls
TrustData
GA4
Other tools
Rolls up into a readiness rung
TrustData
GA4
Other tools
| Feature | TrustData | Google Analytics 4 | Other tools |
|---|---|---|---|
| Published numeric bands with floors and ceilings | |||
| Server-side fetch, sees what a crawler sees | Sometimes | ||
| Signals differ by site type | |||
| Field-level diff between crawls | |||
| Rolls up into a readiness rung |
FAQ
It scores whether a page can be parsed, chunked and quoted by an AI engine, which is a different question from whether it ranks. TrustData fetches each page server-side, parses the HTML, and checks 13 binary signals along with 11 structural bands covering document architecture, information chunking and visual emphasis. The output is the specific bands the page fell outside, not a grade.
Because both ends fail. Heading density below 0.5 per 100 words means the page has no structure to chunk on; above 3.0 means headings stop carrying meaning. Paragraphs under 30 words fragment an idea across chunks; over 90 they bury the answer. A list ratio under 0.10 means nothing is enumerable, over 0.40 means the prose has become a bullet dump with no connective reasoning. Every band has a floor and a ceiling for that reason.
They derive from published experimental work on generative-engine extraction, and they are deliberately treated as a first pass rather than a law. The intention is to regress them against observed citation rates on real properties, so the bands tighten as evidence accumulates. Until then they are documented rather than hidden, which is the difference between a threshold you can argue with and a black-box score you cannot.
Yes. Beyond the shared signals, a lead-generation site is checked for comparison content, testimonials, case studies, use cases and social proof, because those are what an engine looks for when answering an evaluation question about a service. An e-commerce site is checked for product schema, visible pricing and reviews instead. Scoring both against the same list would penalise one for lacking something it should not have.
The average audit score across your pages is the Be Retrieved rung, which is third. It sits above Get Found, because a page a crawler cannot reach cannot be extracted whatever its structure, and below Be Trusted, because a perfectly structured page still needs corroboration to be cited with confidence. If Get Found is failing, the audit is not your bottleneck yet, and the ladder will say so by marking the lower rung as the focus.
14-day free trial
14-day free trial. Page audits are part of AI Visibility, included in every plan.